Anmerkungen:
Nach Informationen von SSRN wurde die ursprüngliche Fassung des Dokuments January 22, 2015 erstellt
Beschreibung:
We investigate the computational complexity of estimating quantile based risk measures, such as the widespread Value at Risk, via nested Monte Carlo simulations. The estimator is a conditional expectation type estimator where two-stage simulations are required to evaluate the risk measure: an outer simulation is used to generate risk-factor scenarios that govern prices and an inner simulation is used to evaluate the future portfolio value based on each scenario. We propose a new set of non-uniform algorithms to evaluate risk. The algorithms place more importance upon outer scenarios which are more likely to have a direct impact on the estimator and considers the marginal changes in the risk estimator at each additional inner scenario. We demonstrate using experimental settings that our proposed algorithms outperform the uniform algorithm and results in a lower variance and bias with the same initial settings and resources. The results are also robust enough for the multidimensionality of risk factors and the non-linearity of pay-offs